dorsal/arxiv
View SchemaSafe Heterogeneous Multi-Agent RL with Communication Regularization for Coordinated Target Acquisition
| Authors | Gabriele Calzolari, Vidya Sumathy, Christoforos Kanellakis, George Nikolakopoulos |
|---|---|
| Categories | |
| ArXiv ID | 2601.08327vv1 |
| URL | https://arxiv.org/abs/2601.08327 |
| License | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ |
Abstract
This paper introduces a decentralized multi-agent reinforcement learning framework enabling structurally heterogeneous teams of agents to jointly discover and acquire randomly located targets in environments characterized by partial observability, communication constraints, and dynamic interactions. Each agent's policy is trained with the Multi-Agent Proximal Policy Optimization algorithm and employs a Graph Attention Network encoder that integrates simulated range-sensing data with communication embeddings exchanged among neighboring agents, enabling context-aware decision-making from both local sensing and relational information. In particular, this work introduces a unified framework that integrates graph-based communication and trajectory-aware safety through safety filters. The architecture is supported by a structured reward formulation designed to encourage effective target discovery and acquisition, collision avoidance, and de-correlation between the agents' communication vectors by promoting informational orthogonality. The effectiveness of the proposed reward function is demonstrated through a comprehensive ablation study. Moreover, simulation results demonstrate safe and stable task execution, confirming the framework's effectiveness.
{
"annotation_id": "0c119c2e-9f0f-424e-be06-c053081409aa",
"date_created": "2026-02-17T05:53:15.167000Z",
"date_modified": "2026-02-17T05:53:15.167000Z",
"file_hash": "0c57dc160f8cfaab95d65caee3d0a33f46b4094513973be1b0360f9731724a6d",
"private": false,
"record": {
"abstract": "This paper introduces a decentralized multi-agent reinforcement learning framework enabling structurally heterogeneous teams of agents to jointly discover and acquire randomly located targets in environments characterized by partial observability, communication constraints, and dynamic interactions. Each agent\u0027s policy is trained with the Multi-Agent Proximal Policy Optimization algorithm and employs a Graph Attention Network encoder that integrates simulated range-sensing data with communication embeddings exchanged among neighboring agents, enabling context-aware decision-making from both local sensing and relational information. In particular, this work introduces a unified framework that integrates graph-based communication and trajectory-aware safety through safety filters. The architecture is supported by a structured reward formulation designed to encourage effective target discovery and acquisition, collision avoidance, and de-correlation between the agents\u0027 communication vectors by promoting informational orthogonality. The effectiveness of the proposed reward function is demonstrated through a comprehensive ablation study. Moreover, simulation results demonstrate safe and stable task execution, confirming the framework\u0027s effectiveness.",
"arxiv_id": "2601.08327",
"authors": [
"Gabriele Calzolari",
"Vidya Sumathy",
"Christoforos Kanellakis",
"George Nikolakopoulos"
],
"categories": [
"cs.RO",
"cs.AI"
],
"license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
"title": "Safe Heterogeneous Multi-Agent RL with Communication Regularization for Coordinated Target Acquisition",
"url": "https://arxiv.org/abs/2601.08327",
"version": "v1"
},
"schema_id": "dorsal/arxiv",
"source": {
"execution_id": "0bd7f945-3567-4a94-9865-8eedd7009fa9",
"id": "arXiv Dataset",
"type": "Model",
"variant": "snapshot-2026-01-17",
"version": "0.1.0"
},
"user_id": 1000002
}